Re: [PATCH 2/7] drm/amdgpu/vce: Align VCPU BO to nearest power of two

["Christian König" <christian.koenig@amd.com>]

On 4/20/26 14:10, Timur Kristóf wrote:
> VCE accesses memory, including its firmware, through a BAR.
> It works slightly differently on each generation.
> In case of VCE1, the start address of this BAR is zero
> and we can't change it due to the firmware validation mechanism.
> 
> Align the VCE VCPU BO to the nearest power of two after the
> byte size of the BO in order to prevent it from crossing the
> boundaries of its BAR.
> 
> Fixes: d38ceaf99ed0 ("drm/amdgpu: add core driver (v4)")
> Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
> index efdebd9c0a1f3..0658f13f0bdf7 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
> @@ -218,7 +218,8 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
>  	if (!adev->vce.fw)
>  		return -ENOENT;
>  
> -	r = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
> +	r = amdgpu_bo_create_kernel(adev, size,
> +				    roundup_pow_of_two(size),
>  				    AMDGPU_GEM_DOMAIN_VRAM |
>  				    AMDGPU_GEM_DOMAIN_GTT,

I don't think that this is a good idea. The VRAM BO should never cross a BAR boundary.

What could be is that the GTT BO does that, but approach that is only available on VCE4 anyway.

I think we should rather restrict falling back to GTT to VCE4 only.

Regards,
Christian.

>  				    &adev->vce.vcpu_bo,